CtrlK
BlogDocsLog inGet started
Tessl Logo

scgpt

Embed and annotate single-cell expression data with scGPT, a foundation model for single-cell biology. Use this skill when: (1) Producing cell embeddings from an AnnData for clustering/integration, (2) Zero-shot or fine-tuned cell-type annotation, (3) Gene-level representation for perturbation/GRN tasks. For probabilistic single-cell models (scVI etc.), use the scvi-tools library.

71

Quality

89%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

86%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A tight, highly actionable skill body with executable examples and a well-sequenced async compute workflow. The main gap is a missing explicit validation step on the remote-compute output before downstream use.

Suggestions

Add an explicit validation checkpoint after the compute_done notification — e.g. assert embedded.h5ad has the expected cell count and an X_scGPT obsm key before saving/using it.

Tighten the remote-compute prose: drop the inline 'cell ends here — kernel never blocks on compute' comment and fold the attach/close pattern into a single concise step.

Consider a one-line 'Prerequisites' note on how to obtain the checkpoint directory (the Drive link) since the body references /path/to/scgpt-human without sourcing it.

DimensionReasoningScore

Conciseness

Lean and mostly efficient — no padding about what single-cell biology or scGPT is — but the remote-compute prose and inline comments (e.g. 'cell ends here — kernel never blocks on compute') could be trimmed slightly.

4 / 5

Actionability

Fully executable, copy-paste-ready code throughout: GeneVocab.from_file, embed_data with real parameters, a complete submit_job call, and a troubleshooting table with a concrete manifest-casing normalization fix.

5 / 5

Workflow Clarity

The remote-compute flow is clearly sequenced (submit → wait_for_notification → act on payload → attach for full result) with a notification feedback loop, but lacks an explicit output-validation step (e.g. verify embedded.h5ad shape/keys).

4 / 5

Progressive Disclosure

Well-organized into clear sections (Prerequisites, How to run, Output format, Remote compute, Gotchas, Troubleshooting) with one-level-deep pointers to related skills; no bundle files are needed and nothing is over-nested.

5 / 5

Total

18

/

20

Passed

Description

92%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, specific description with an explicit 'Use when' trigger list and a helpful boundary against the related scvi-tools niche. Trigger-term coverage is good but could add a couple of common synonyms/extensions a user might naturally say.

DimensionReasoningScore

Specificity

Lists multiple concrete actions — 'Producing cell embeddings', 'cell-type annotation', 'Gene-level representation for perturbation/GRN tasks' — covering the model's main capabilities comprehensively.

5 / 5

Completeness

Explicitly answers both what ('Embed and annotate single-cell expression data with scGPT') and when ('Use this skill when: (1)... (2)... (3)...') with concrete numbered trigger phrases.

5 / 5

Trigger Term Quality

Good natural-domain coverage ('single-cell', 'embeddings', 'clustering/integration', 'cell-type annotation', 'perturbation/GRN'), but a few common variants/extensions (e.g. 'AnnData', '.h5ad') a user might say are absent from the description.

4 / 5

Distinctiveness Conflict Risk

Clear niche (scGPT foundation model for single-cell biology) with an explicit de-confliction boundary — 'For probabilistic single-cell models (scVI etc.), use the scvi-tools library' — minimizing wrong-skill triggering.

5 / 5

Total

19

/

20

Passed

Validation

81%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation13 / 16 Passed

Validation for skill structure

CriteriaDescriptionResult

metadata_version

'metadata.version' is missing

Warning

metadata_field

'metadata' should map string keys to string values

Warning

frontmatter_unknown_keys

Unknown frontmatter key(s) found; consider removing or moving to metadata

Warning

Total

13

/

16

Passed

Repository
UnicomAI/wanwu
Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.